The Signature algorithm was designed as part of a diploma thesis. The idea behind this algorithm is to approximate gestures with a signature whereas the signature is computed from example gestures as done in the Rubine algorithm. The signatures are compared based on distance functions leading to a classification.
For the creation of the signature we use a grid which consists of squares of the same size. Each square is identified with a bit string and two neighbouring squares always differ in exactly one bit.
During the preprocessing phase the gestures are stretched to a uniform size and mapped onto the grid. Each point of the gesture can now be represented with the bit string of its related square as shown in Figure 4.2. The full signature consists of the concatenation of these bit strings.
|GRID_SIZE||The grid size defines the number of cells the grid should have within a line.|
|RASTER_SIZE||The raster size defines the width of and height a gesture is stretched to.|
|DISTANCE_FUNCTION||The full qualified class name of the distance function. The class has to implement the DistanceFunction interface.|
<configuration> <algorithm name="org.ximtec.igesture.algorithm.signature.SiGridAlgorithm"> <parameter name="DISTANCE_FUNCTION">org.ximtec.igesture.algorithm.signature.HammingDistance</parameter> <parameter name="RASTER_SIZE">120</parameter> <parameter name="MIN_ACCURACY">0.5</parameter> <parameter name="GRID_SIZE">8</parameter> </algorithm> </configuration>